The Agent Skills Directory

[DYNAMIC_EXECUTION]: The script scripts/office/soffice.py dynamically generates and compiles a C-based socket shim at runtime using gcc. This shim is loaded via LD_PRELOAD to enable LibreOffice functionality in environments where AF_UNIX sockets might be restricted. This constitutes runtime code generation and library injection.
[COMMAND_EXECUTION]: Several scripts utilize the subprocess module to execute system commands and external tools:
scripts/office/soffice.py executes gcc for shim compilation and soffice for document processing.
scripts/thumbnail.py executes pdftoppm (or optionally uses pdf2image) for slide-to-image conversion.
scripts/office/validators/redlining.py executes git diff for comparing document versions.
[INDIRECT_PROMPT_INJECTION]: The skill provides a large attack surface for indirect prompt injection by processing external user-supplied PowerPoint files.
Ingestion points: Slide content is extracted and processed in scripts/office/unpack.py and scripts/thumbnail.py.
Boundary markers: Absent. No explicit delimiters or instructions are provided to the agent to distinguish between slide content and operational instructions.
Capability inventory: The skill has extensive capabilities, including file system modification, command execution, and runtime compilation, which could be leveraged if an injection is successful.
Sanitization: The skill uses defusedxml for XML parsing, which effectively mitigates XML External Entity (XXE) attacks, but it does not sanitize or filter the textual content of the slides.

pptx